PrimeGrid Races 2018

[TennesseeTony]

In the rare event I get to shovel snow....I'm wearing a t shirt....and soaked in sweat. (But I'm only slightly fat)

 

[Howdy]

In the rare event I get to shovel snow....I'm wearing a t shirt....and soaked in sweat. (But I'm only slightly fat)

I shovel in a sweat shirt, same as yourself I sweat. Tried a T-shirt before was a little too cold for my old bones. I shoveled yesterday and woke up today with ~3" more to shovel again......

 

[Howdy]

Well I'm out early on this, seems most of us are weeding out the projects that are not primes (yes I found 2). Thus making it easier for the big guns to sweep them up, turning towards something more meaningful until FB starts up. Would be nice if more people were able to find some too.

 

[ao_ika_red]

16-23 March 18:00:00 321-LLR Year of the Dog Challenge 7 days

CPU app only?

 

[StefanR5R]

Yes.

and of course:

 

[ao_ika_red]

Thanks. Luckily it won't overlap so much with FB sprint.

 

[StefanR5R]

Everybody still has got an app_config.xml left from the Number Theory Week Challenge in September 2017. But for convenience, here it is again:

<app_config>
    <app_version>
        <app_name>llr321</app_name>
        <cmdline>-t 4</cmdline>
        <avg_ncpus>4</avg_ncpus>
    </app_version>
</app_config>

Replace the two occurrences of "4" by what you think, or measured, to be best on your host.

I read somewhere that the <fraction_done_exact/> block which is still seen in various example app_config.xml's is no longer necessary.

The example in this challenge's announcement post also includes a <max_concurrent> tag, which you may or may not find useful. It of course controls the maximum number of llr321 tasks to run in parallel.

 

[ao_ika_red]

This will be a nice buildup for upcoming YoTD event. Thanks Stefan.

 

[Ken g6]

Everybody still has got an app_config.xml left from the Number Theory Week Challenge in September 2017. But for convenience, here it is again:

<app_config>
    <app_version>
        <app_name>llr321</app_name>
        <cmdline>-t 4</cmdline>
        <avg_ncpus>4</avg_ncpus>
    </app_version>
</app_config>

Replace the two occurrences of "4" by what you think, or measured, to be best on your host.

I read somewhere that the <fraction_done_exact/> block which is still seen in various example app_config.xml's is no longer necessary.

The example in this challenge's announcement post also includes a <max_concurrent> tag, which you may or may not find useful. It of course controls the maximum number of llr321 tasks to run in parallel.

It's even better if you set avg_ncpus to 1 and set your BOINC options to use only one CPU.

 

[crashtech]

It's even better if you set avg_ncpus to 1 and set your BOINC options to use only one CPU.

Would you elaborate?

 

[Ken g6]

Would you elaborate?

<app_config>
    <app_version>
        <app_name>llr321</app_name>
        <cmdline>-t 4</cmdline>
        <avg_ncpus>1</avg_ncpus>
    </app_version>
</app_config>

Then set the percentage of cores used so it only means to use one core.

Actually this doesn't matter that much if you don't need to return the WUs immediately. But it might keep the timing results consistent.

 

[crashtech]

Unfortunately I am still confused. I thought multithreading was good, up to a point.

 

[Ken g6]

"-t 4" or whatever number makes LLR do multithreading. Setting avg_ncpus doesn't prevent that; it just tells BOINC to act like it's not multithreaded. But it is.

Still confused? Well, BOINC is confused by multithreading, so this makes it less confused even if you're more confused.

 

[crashtech]

Hmm, I guess I'll have to play with it and see what it does. Thanks.

 

[StefanR5R]

@Ken g6, thanks! (Some of my hosts will have to run several multithreaded LLR processes at once though.)

@crashtech, remember, <cmdline> influences the science application (and only the science application), whereas <avg_ncpus> influences boinc-client's task scheduler (and only the client's task scheduler) --- at PrimeGrid. (On some other projects, <avg_ncpus> is picked up by the science application. By Cosmology@Home's vboxwrapper based applications for example.)

Then set the percentage of cores used so it only means to use one core.

This too only influences the client's task scheduler. An easy way to set the percentage to mean 1 logical CPU is to set 1 %. This is rounded up to 1 logical CPU on hosts which have fewer than 100 logical CPUs...

--------
Yesterday I finally started looking into benchmarking PrimeGrid's LLR application by running it standalone with the same WU for each test run with different -t parameters. (IOW, running it outside of the boinc client, and not with random ever-varying WUs.) Will post when I got this testing method figured out.

 

[Howdy]

I just wait until someone posts a config and copy and paste.........change the numbers to what looks like they work and run...........probably not the most efficient method but it works for me!! Or I end up with this:

<app_config><you have no clue>
<app_version><yeah right_stick to copy and paste>
<app_name>llr321</app_name><pffft_what's in a/name?/>
<cmdline>-t 4</cmdline><don't you EVEN try/to_tell ME what to do!?!/>
<avg_ncpus>1</avg_ncpus><AVG?AVG?run all dem B!t%^$>
</app_version><the copy and paste one Remember?>
</app_config><still not a clue>

 

[StefanR5R]

I tried your config but suddenly my boinc plays Tetris instead of looking for primes! Help!

 

[Howdy]

I tried your config but suddenly my boinc plays Tetris instead of looking for primes! Help!

Thank you I needed a laugh after the past few days!!!

 

[ao_ika_red]

This one will start at 18:00 UTC, right? Then, I need some advice to let BOINC client download llr321 tasks automatically because it will be pretty late around here. Will "daily schedules" tab on BOINC's "computing preferences" do the job?

 

[Ken g6]

Yes, that's a good way to do it. Adjust your networking schedule so it doesn't download anything between now and then.

 

[Howdy]

PG challenges involving only CPU SUCK!!! That is all TYVM

(I just need some CPU power..........thus my machines suck!! )

 

[ao_ika_red]

Yes, that's a good way to do it. Adjust your networking schedule so it doesn't download anything between now and then.

Uh la la, it kept downloading task even though the networking schedule is already set. I think I will do it manually as my head is getting dizzy.

 

[StefanR5R]

The networking schedule is supposed to work. And I had it working for me in the past.

However, this challenge is 168 hours long. If you miss the first 6 or 7 hours, you are still in for more than 95 % of it.

Edit,
you are located at UTC+07:00, right? Which puts the beginning of the challenge at Saturday 01:00 AM of your local time. Then you need to set a networking schedule of

Friday 00:00 - 01:23 (for example) and
Saturday 01:01 - 24:00​

IOW you need to prevent downloads on Friday and allow downloads on Saturday (after the start). And of course remember to remove the schedule later on tomorrow, or at least before next Friday.

 

[Markfw]

PG challenges involving only CPU SUCK!!! That is all TYVM

(I just need some CPU power..........thus my machines suck!! )

I will have a 6700k on ebay soon. motherboard/cpu and maybe ram. PM me if interested.

 

[zzuupp]

& off we go!